California rolls out new rules to trace AI-generated images and videos

California has begun enforcing Senate Bill 942 (SB 942), the California AI Transparency Act, requiring major generative AI providers to attach hidden digital provenance information to AI-generated images, videos and recordings in an effort to combat deepfakes.

The disclosures can reveal which AI system created or altered content and when, while free detection tools will allow users to check compatible files.

Beginning in 2027, large platforms will have to preserve and surface available provenance data, followed by requirements for new recording devices in 2028.

California sets $5,000 penalties for AI transparency violations

California’s SB 942, known as the California AI Transparency Act, establishes disclosure and provenance requirements for major generative AI providers.

Covered providers must let users add clear, visible labels identifying AI-generated image, video and audio content, while their systems must embed permanent or difficult-to-remove hidden disclosures containing information such as provider, model and version, creation or alteration time and a unique identifier.

The law also requires providers to make free AI detection tools publicly available, including through an API, while restricting the collection and retention of users’ personal information and submitted content.

Providers licensing their AI systems to third parties must ensure those systems retain provenance capabilities and revoke licenses within 96 hours if a licensee disables them.

Noncompliance carries a $5,000-per-violation civil penalty.